基于生态地理分区的5套土地利用/覆盖数据的不确定性研究  

Study on the Uncertainty of Five Land Use/Land Cover Data Sets based on Eco-graphical Regions

作　　者：王洁[1,2] 张增祥[1] 张委伟[1,2] 

机构地区：[1]中国科学院遥感应用研究所,北京100101 [2]中国科学院大学,北京100049

出　　处：《遥感技术与应用》2012年第6期865-872,共8页Remote Sensing Technology and Application

基　　金：国家海洋局"渤海环境立体监测与动态评价专项"渤海环境遥感监测技术开发和业务化应用专题(BH2009RS)资助

摘　　要：基于生态地理分区从面积精度和位置精度两个方面定量探讨了5套全球土地利用/覆盖(LULC)数据产品在实际应用中的不确定性,为基于生态地理分区的相关研究选择合理数据集提供参考依据,同时为中国生产全球LULC产品提供有关信息。选择中国地区20个典型生态地理分区为研究对象,采用最小误差频率法,分析5套数据集对各个类型面积估计的不确定性大小及原因;采用混淆矩阵法,基于位置分析5套数据集在类型混分方面的不确定性大小,原因及空间分布规律。结果表明:在生态地理分区尺度,MODIS,Meris300以及Glc2000这3套数据集明显优于Umd和Usgs这两套数据集,并且随着生态地理分区自南向北、自东向西的空间分布,这3套数据集的不确定性呈减小趋势。对所有生态地理分区而言,Meris300数据集整体估计的稳定性最高,但是估计精度不是最高,并且它对建设用地和水域的估计最有优势。MODIS数据集整体估计精度和稳定性次之,对耕地的估计最有优势。Glc2000数据集更适用于土地利用/覆盖简单的生态地理分区。研究还发现地形和土地利用/覆盖的复杂程度是引起数据集不确定性的两个重要因素。Based on eco-geographical regions, this paper, from two aspects of area accuracy and position accuracy, quantitatively studied the uncertainty of five global Land Use and Land Cover（LULC）data sets in practical applica tion. The conclusions of this paper can be employed in choosing reasonable data sets efficiently for other relative re searches about eco-geographical regions. In addition, the results will also provide reference for Chinese researchers to make global LULC map. In this paper,twenty typical eco-geographical regions were chosen as research objects. On the one hand,in the estimations of thematic type area,the uncertainty degree and reasons of each data set were analyzed through counting the minimum error of each data set; On the other hand, error matrix was adopted to study the position uncertainty of each LULC type in five data sets,reasons and spatial distribution laws. The results indicated that,in eco-geographical regions scale, these three data sets of MODIS,Meris300 and Glc2000 were super to Umd and Usgs. And for the distribution of the eco-geographical regions from south to north,from east to west, the uncertainty of these three data sets would be decreased gradually Meanwhile,Merisg00 data set estimated the area and position of each LULC type with the highest stability,while the accuracy was not the highest. It was good at the estimation of construction land and water. Both estimation stabiiity and accuracy of MODIS data set were ac- ceptable and it was used to estimate farmland with lowest error. Glc2000 data set was more suitable to be applied to study the eco-geographical regions with simple LULC type.

关 键 词：生态地理分区 土地覆盖与利用 遥感数据 不确定性 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]